Background
Despite improvement in resuscitation, the neurological outcome of comatose patients after cardiac arrest remains extremely poor
Currently, several clinical parameters and electro-encephalographic (EEG) patterns are recognised as being strongly associated with a poor outcome in unsedated comatose survivors of cardiac arrest; these include absence of pupillary or corneal reflexes, absence of extensor motor response to pain 3 days after cardiac arrest, myoclonus or epilepticus status within the first day after resuscitation, and a burst-suppression or isoelectric EEG pattern
In this context, the serum Neuron-Specific Enolase (NSE), a biomarker of hypoxic brain damage which can be measured easily and reproducibly with minor invasiveness in patients, has recently been assessed as a prognostic predictor after cardiac arrest in several studies
Therefore, we conducted a prospective cohort study to assess the capability of NSE, measured at fixed times, to predict a poor outcome (death or permanent vegetative state) with certainty in a predefined post cardiac-arrest comatose population.
Methods
Patients
We conducted a prospective cohort study of all consecutive non trauma patients who suffered out-of- or in-hospital cardiac arrest and were subsequently admitted to the adult intensive care unit in the Caen University Hospital and the Saint Lô General Hospital from June 2007 to February 2009. Patients who died or awoke within the first 48 hours of admission were excluded from this analysis. Therefore, only patients who remained in coma at 48 hours after cardiac arrest were included in the analysis. All patients were followed for 3 months after cardiac arrest or until death.
This study was submitted to the local ethics committee. The ethical board decided that approval was not necessary given the observational nature of this prospective study. Thus, in accordance with French legislation at the time of the study, no informed consent was obtained from the patients.
Assessment of outcome
Neurological status at 3 months was assessed by telephone interview for patients discharged alive from the intensive care, using the 5-grade Glasgow-Pittsburgh Cerebral Performance Category (GP-CPC) scale
Measurement of serum NSE
Blood samples were collected at 24 h and 72 h after the time of cardiac arrest. All samples with visible hemolysis were discarded from analysis to avoid any falsely elevated values for serum NSE. Blood was centrifuged at 3 000 rpm for 10 min. The isolated serum was immediately frozen at -80°C and stored until time of assay. The serum NSE level was measured using a solid-phase immunoassay with double monoclonal antibodies directed against NSE (Roche Diagnostics GmbH, Mannheim, Germany) on an Elecsys instrument. The limit of detection was 0.05 ng/mL and the institutional normal value was < 16.3 ng/mL. When the NSE level reached 50 ng/mL, the serum was diluted to avoid a hook effect. Clinicians were blinded to NSE results during the entire patient stay. For each patient, the highest measurement of NSE was tested for outcome prediction. Because the 24 h value is more variable than that at 72 h, we also considered the 72 h NSE value as the highest observed NSE in the sensitivity analysis.
Data collection
Clinical variables collected at baseline were: age, sex, underlying diseases, cause of the arrest (cardiac, respiratory, other or unknown), time between arrest and cardiopulmonary resuscitation, initial cardiac rhythm (ventricular fibrillation or tachycardia, asystole, pulseless rhythm), duration of cardiopulmonary resuscitation, number of external electric shocks, cumulative epinephrine dose, and scoring of disease severity within the first day in ICU as assessed during admission by the Simplified Acute Physiology Score type II (SAPS II)
Following our standard of treatment
Clinical parameters included pupillary light reflex (present/absent), motor response to painful stimulation (extensor or absent response/other response), corneal reflex (present/absent), tonic-clonic seizures (present/absent) and myoclonus (present/absent).
Electrophysiological assessment included EEG and Somatosensory-Evoked Potential (SSEP) recordings, routinely performed in our centre. All EEGs and SSEPs were read by an expert neurophysiologist (O.E.). Two EEG were performed: the first within 24 h-36 h after resuscitation and the second at 72 h. EEGs were recorded on a system with at least 10 channels and needle electrodes and used a 10-20 international system (Fp1, Fp2, C3, C4, T5, T6, O1, and O2). The EEG patterns were classified according to the classification system of Synek et al.
Treatment and treatment restriction
All patients were actively supported during the first days following cardiac arrest or until SSEP assessment. They received standard intensive care management and monitoring. In addition, in our practice, therapeutic hypothermia (target temperature 33°C) was recommended for all cardiac causes of arrest, regardless of initial rhythm, and left to the assessment of the attending physician for the other causes. The patients who underwent cooling received propofol or midazolam and sufentanyl for sedation and atracurium (0.5 mg/kg per hour) to prevent shivering. These drugs were stopped after passive rewarning to a central temperature of 36°C. Hypothermia was induced using an endovascular cooling catheter (IcyTM, Alsius, Irvine, CA, USA) inserted into the inferior vena cava via the femoral vein and connected to a cooling device (Coolgard 3000TM, Alsius, Irvine, CA, USA), and was maintained for 24 h. In patients with a bilateral lack of cortical response (N20) to SSEPs, further treatment was considered futile and active care was withdrawn. In addition, in accordance with our previous report
Statistical analysis
Categorical variables were reported as counts and percentages. Quantitative data which did not follow a Gaussian curve (NSE levels) were described as median and Interquartile Range [IQR, first quartile - third quartile]. Other qualitative data were reported as mean ± standard deviation (SD). Means were compared between two groups using Student's
Results
Baseline characteristics and neurological outcome
Of the 149 consecutive patients resuscitated after a cardiac arrest and admitted to the intensive care unit during the study period, 97 fulfilled inclusion criteria for analysis. At 3 months 72 (74%) patients had a poor outcome (CPC 4-5) and 25 (26%) had a good outcome (CPC 1-2, n = 17 and CPC 3, n = 8) as shown in Figure
Profile of the study
Profile of the study.
Patient baseline characteristics are presented in Table
Baseline characteristics
All patients
n = 97
Poor outcome
n = 72
Good outcome
n = 25
p
Age (years), mean ± SD
57 ± 16
60 ± 15
50 ± 17
0.01
Male, n (%)
75 (77)
56 (78)
19 (76)
0.85
Medical history n (%)
Neurologic diseases
11 (11)
10 (14)
1 (4)
0.18
Cardiovascular diseases
Ischemic
20 (21)
19 (26)
1 (4)
0.017
Hypertensive
34 (35)
29 (40)
5 (20)
0.067
Congestive
12 (12)
10 (14)
2 (8)
0.44
- Arrhythmic
4 (5)
3 (4)
2 (8)
0.82
Metabolic diseases
Diabetes mellitus
17 (18)
17 (24)
0
0.007
Respiratory diseases
COPD
7 (7)
6 (8)
1 (4)
0.47
Liver diseases
Cirrhosis
4 (4)
4 (6)
0
0.53
-
Resuscitation variables
Witnessed CA n (%)
77 (79)
57 (79)
20 (80)
0.93
In-hospital CA n (%)
30 (31)
26 (36)
4 (16)
0.06
Primary cause of CA n (%)
Cardiac
55 (57)
38 (53)
17 (68)
0.19
Respiratory
17 (18)
16 (22)
1 (4)
0.04
Other or unknown
22 (23)
17 (24)
5 (20)
0.7
Time from CA to CPR (minutes)
5.6 ± 6.9
6.3 ± 7.2
3.4 ± 5.5
0.07
<3 n (%)
46 (51)
31 (46)
15 (65)
3 -5 n (%)
9 (10)
6 (9)
3 (13)
>5 n (%)
36 (40)
31 (46)
5 (22)
Duration of CPR (minutes)
24.9 ± 24.2
25.7 ± 26.6
22.5 ± 15.7
0.39
<5 n (%)
9 (9)
5 (7)
4 (16)
5 -15 n (%)
18 (19)
13 (18)
5 (20)
>15 n (%)
69 (72)
53 (75)
16 (64)
Primary rhythm n (%)
0.09
Asystole
55 (57)
46 (64)
9 (36)
VF/VT
35 (36)
21 (29)
14 (56)
Pulseless electrical activity
5 (5)
4 (6)
1 (4)
Unknown
2 (2)
1 (1,4)
1 (4)
Number of defibrillations, mean ± SD
2.1 ± 3.1
1.9 ± 3.3
2.8 ± 2.4
0.17
Epinephrine mg mean, ± SD
5.1 ± 5.2
5 ± 4.8
5.6 ± 6.2
0.62
ICU admission
SAPS II mean, ± SD
67 ± 15
69 ± 15
61 ± 11
0.006
Shock n(%)
59 (61)
43 (60)
16 (64)
0.7
Renal replacement therapy n(%)
15 (16)
11(15)
4 (16)
1
MOF n(%)
18 (19)
16 (22)
2 (8)
0.14
Therapeutic hypothermia n(%)
65 (67)
45 (63)
20 (80%)
0.11
CA, cardiac arrest; COPD, chronic obstructive pulmonary disease; CPR, cardiopulmonary resuscitation; FV ventricular fibrillation; MOF, multi organ failure according to Kraus criteria
Serum NSE levels and prediction of poor outcome
The median levels of NSE at 24 h (n = 86) and at 72 h (n = 61) and the highest measurement of NSE for each patient were significantly higher in patients with poor outcomes: median 24 h NSE level: 59.4 ng/mL [IQR, 37-106] versus 28.8 ng/mL [IQR, 18-41] (
Frequency of different neurological outcome categories in relation to the individual highest measurement of NSE
Frequency of different neurological outcome categories in relation to the individual highest measurement of NSE.
In our sample group, the highest measurements of NSE for each patient were not different in patients treated with or without induced hypothermia: 67.4 ng/mL [IQR, 37.2-143.6] versus 55.3 ng/mL [IQR, 25.1-159.4] (
The ROC curve for poor outcome for the highest measurement of NSE for each patient is presented in Figure
Receiver operating characteristic curves for different peak serum NSE cut-off values (ng/mL) to predict poor neurological outcome
Receiver operating characteristic curves for different peak serum NSE cut-off values (ng/mL) to predict poor neurological outcome.
The sensitivity analysis provided similar results (see additional file
" Receiver operating characteristic curves for 72 h NSE values (ng/mL) to predict poor neurological outcome. This file highlight that considering 72 h-NSE values (n = 61), a level ≥ 68 ng/mL predicted a poor outcome (CPC 4-5) with a positive predictive value of 100% [95% IC = 100% - 100%] and a sensitivity of 67% [95% IC = 54% - 81%].
Click here for file
Contribution of each clinical, electrophysiological and biological test to prediction of poor outcome
The absence of cortical response to SSEPs was recorded in 45 patients. An unfavorable SSEP result was associated with 2.9 ± 1 predefined unfavorable clinical-EEG criteria. In accordance with our treatment restriction policy, all died.
The predictive values of clinical-EEG and biological tests for a poor outcome are presented in Table
Prediction of poor outcome with clinical, electrophysiological and biological variables
Patients tested, n
Abnormal test result, % (95% CI)
False positive rate*, % (95% IC)
Positive likelihood ratio (95% IC)
Motor response ≤ 2
- 24 h-36 h
97
87% [80% - 93%]
56% [45%- 67%
1.7 [1.2 - 2.5]
- 72 h
87
70% [61% - 80%]
16% [7% - 25%]
5.8 [2.3 -1 4.1]
No corneal reflexes
- 24 h-36 h
95
40% [30% - 50%]
12% [2% - 22%]
4.2 [1.4 - 12.4]
- 72 h
86
36% [26% - 46%]
0% [0%-14%]
26 [1.7 - 415]
No pupillary reflexes
- 24 h-36 h
97
26% [17% - 35%]
4% [4% - 12%]
8.3 [1.2 - 58]
- 72 h
87
21% [12% - 29%]
0% [0% - 14%]
15 [1 - 244]
Myoclonus
- 24 h-36 h
97
27% [18% - 36%]
0% [0% -14%]
19 [1.2 - 299]
- 72 h
87
23% [14% - 32%]
8% [4% - 20%]
3.6
Epilepsy
- 24 h-36 h
97
7% [7% - 12%]
0% [0% -14%]
5.3 [0.3-90]
- 72 h
87
7% [2% - 12%]
0% [0% -14%]
5.3 [0.31-91]
- Malignant EEG **
- 24-36 h
90 ¶
44% [34% - 55%]
4% [2% - 10%]
14 [2.1-97]
- 72 h
73¶¶
44% [33% - 55%]
0% [0% -17%]
25 [1.6-394]
Highest measurement of NSE ≥ 47 ng/mL
97
58% [48% - 68%]
7% [0,4% - 14%]
4.5 [1.8 -11]
Highest measurement of NSE ≥ 97 ng/mL
97
36% [27% - 46%]
0% [0% -13%]
25 [1.6-397]
* Patients with abnormal test results and
- good
** included isoelectric and burst-suppression pattern with interburst interval of at least 1s and generalised continuous epileptiform discharges
¶ Forty-four patients were still affected by sedative medication
¶¶ twenty-one patients were still affected by sedative medication
An approach based on a combination of SSEPs, NSE and clinical-EEG tests increased the number of patients (63/72 (88%)) identified as having a poor outcome, as shown in Figure
Predictors of poor outcome according to different clinical-EEG, NSE and SSEPs combinations
Predictors of poor outcome according to different clinical-EEG, NSE and SSEPs combinations. * included: myoclonus at 24 h, and absence of pupillary light reflex or corneal reflex, tonic-clonic seizures and malignant EEG pattern at 72 h.
No tests predicted a good outcome. For example, only 51% (24/47) of our patients with a favorable SSEP result made a good recovery
Discussion
Considering the poor neurological prognosis of comatose patients after cardiac arrest, physicians are rapidly confronted with the ethical question of whether to continue intensive treatment. To our knowledge, except for unfavorable SSEP results, predictors of poor outcome with a 100% specificity and a high sensitivity are lacking
In our study, 74% of the patients who remained comatose after cardiac arrest never regained consciousness. This result is consistent with previous studies
NSE levels were significantly higher in comatose cardiac arrest survivors with poor outcome, regardless of the time of measurement. In addition, we identified a cut-off value for NSE (> 97 ng/mL), predicting a poor outcome (death or vegetative state) with no false positives. Interestingly, 32/35 (91%) and 29/35 (83%) patients with an NSE measurement > 97 ng/mL had an unfavorable SSEP result and a malignant EEG pattern at 72 h, respectively. These findings argue that NSE measurement may be used, in combination with clinical tests, as a potential substitute for SSEPs and EEG in settings without access to electrophysiological assessment, for early identification of a subgroup of irrecoverable patients for whom continued intensive care could be considered futile.
However, the cut-off values for NSE with 100% predictive value of poor outcome determined in our study is rather high in comparison with others (range 9 to 91 ng/mL) previously reported (Table
Comparison of NSE levels to predict poor outcome after cardiac arrest and referenced study profiles
Référence
In-hospital CPR (%)
Time of inclusion after CPR,
n
Hypothermic therapy,
n
Follow-up
Poor outcome definition, and number
n
Method used for NSE measurement
NSE sampling time
Cut-off value (ng/mL)
Se
Sp
Fogel et al.
Not specified
ICU admission
n = 43
No
3 months
Remained comatose
n = 25
Radioimmunoassay, Pharmacia LKB
Day 0
Day 1
Day 2
Day 3
33
33
33
33
25
60
63
65
100
100
100
100
Martens et al.
Not specified
> 24 h
n = 64
No
6 months
Remained comatose
n = 35
Radioimmunoassay, Profilogen
24 h
20
51
89
Schoerkhber et al.
Yes (not specified)
> 6 h
n = 56
No
6 months
CPC 3-5
n = 28
Radioimmunoassay, Profilogen
12 h
24 h
48 h
72 h
Peak NSE level 12-72 h
38
40
25
17
27
18
8
48
70
29
100
100
100
100
100
Rosen et al.
No
> 24 h
n = 66
No
1 year
CPC 3-5
n = 42
Imunoluminometric assay, Byk Sangtec Diagnostica
Day1
Day2
Day3
25
25
25
NM
NM
NM
100
100
100
Zingler et al.
Yes (not specified)
ICU admission
n = 27
No
3 months
CPC 4-5
n = 17
Immunoluminometric assay, Byk Sangtec Diagnostica
Day1
Day2
Day3
Day7
48
43
91
39
53
91
75
57
100
100
100
100
Tiainem et al.
No*
ICU admission
n = 70
Yes
n = 36
6 months
CPC 3-5
n = 29
time-resolved immunofluorometric assay (DELFIA, Wallac)
24 h HT/no HT
36 h HT/no HT
48 h HT/no HT
31/13
26/13
25/9
22/59
30/63
25/76
96/100
96/100
96/100
Meynaar et al.
Yes (23%)
ICU admission
n = 110
No
Hospital
discharge
Remained comatose
n = 81
time-resolved immunofluorometric assay (DELFIA, Wallac)
Peak NSE level 24-48 h¶
25
59
100
Pfeifer et al.
Yes (44%)
> 48 h
n = 97
Not specified
Day 28
CPC 4-5
n = 70
Imunoluminometric assays, Byk Sangtec Diagnostica
Day 3
65
50
96
Rech at al.
Yes (100%)
>12 h
n = 45
No
6 months
CPC 4-5
n = 34
Electrochemiluminescence immunoassay, Roche Mannheim
Between 12-36 h
60
35
100
Zandbergen et al
Not specified
> 24 h
n = 407
Yes (not specified)
1 months
CPC 4-5
n = 356
Immunoluminometric assay, Byk Sangtec Diagnostica
24 h
48 h
72 h
>33
>33
>33
42
52
46
100
100
100
Auer et al.
Yes (not specified)
n = 17
Not specified
Hospital
discharge
Death
n = 9
Electrochemiluminescence immunoassay, Roche Mannheim
48 h
30
79
100
Grubb et al.
No
ICU admission
n = 143
Not specified
Hospital
discharge
Death
Enzyme immunoassay, Roche Diagnostics
12 h
24-48 h
72-96 h
NM
71
NM
14
100
Reisinger et al.
Yes (44%)
ICU admission
n = 177
Yes
n = 20
6 months
CPC 4-5
n = 59
Electrochemiluminescence immunoassay, Roche Mannheim
Peak NSE level
Day 0-4
80
63
100
Oksanen et al.
No*
ICU admission
n = 90
Yes
n = 90
6 months
CPC 3-5
n = 40
Electrochemiluminescence immunoassay, Roche Mannheim
24 h
48 h
41
33
20
43
100
100
Rundgren et al.
Yes (17%)
ICU admission
n = 102
(for NSE cohort)
Yes
n = 102
6 months
CPC 3-5
n = 46
Imunoluminometric assay, DiaSorin
2 h
24 h
48 h
72 h
31
38
28
27
6
11
67
50
100
100
100
100
Shinozaki et al.
Yes (27%)
ICU admission
n = 80
Yes
n = 45
6 months
CPC 3-5
n = 67
Immunoradiometric assay, Profilogen, DiaSorin
Admission
6 h
24 h
46
66
40
14
19
72
100
100
100
Steffen et al
Yes (21%)
ICU admission
n = 240
Yes
n = 133
ICU discharge
CPC 3-5
n = 147
Electrochemiluminescence immunoassay, Roche Mannheim
72 h HT/no HT
79/27
50
80
100
100
Present study
Yes (31%)
> 48 h
n = 97
Yes
n = 65
3 months
CPC 4-5
n = 72
Electrochemiluminescence immunoassay, Roche Mannheim
Peak NSE level 24-72 h
97
49
100
HT; hypothermic therapy, NM; not mentioned
* Only ventricular fibrillation as initial rhythm was eligible for the study
¶ Results mentioned only for patients who remained comatose 48 hours after CPR (n = 67)
Whether hypothermia, performed in most of our patients, may affect NSE measurement is unclear. Recently, studies
Therefore, considering decisions to continue treatment in comatose patients after cardiac arrest, the different cut-off values for NSE reported in the literature should be interpreted with caution as many patient, treatment, and assay-procedure related factors may influence NSE measurement. For example, if we had used the NSE threshold > 33 ng/mL, defined by Zandbergen et al.
We also report that an approach based on a combination of electrophysiological, biological and clinical tests allowed to increase the number of patients identified as having a poor outcome (Figure
This study has some limitations. Firstly, the relative small sample size may limit the interpretation and relevance of the cut-off value for NSE. In addition, because of differences in NSE measurements obtained using assays from different manufacturers (Table
Conclusion
We show that NSE levels, measured early in the course of patient care for those who remained comatose 3 days after cardiac arrest, are significantly higher in patients with a poor outcome (death or vegetative state). In addition, we provide a cut-off value (> 97 ng/mL) for NSE, with 100% predictive value for a poor outcome. Nevertheless, considering decisions to continue treatment in this setting, we emphasize that cut-off values for NSE presented in the literature should be interpreted with caution and that an approach based on a combination of SSEPs, NSE and clinical-EEG tests would be the most accurate for early identification of a subgroup of irrecoverable patients for whom intensive treatment could be regarded as futile and palliative care only could be provided.
Abbreviations
EEG: Electro-Encephalography; GPCPC: Glasgow-Pittsburgh Cerebral Performance Category; NSE: Neuron-Specific Enolase; SSEPs: Somatosensory Evoked Potentials; ROC: Receiver Operating Characteristics.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
CD and CQ initiated the study, and the design. SA performed NSE measurement. OE read all EEGs and SSEPs. JJP, CG and CD performed the statistical analysis and were involved in the interpretation of the results. CD and CQ wrote the manuscript, and JJP and PC helped to draft the manuscript. AS, XV, FP, MR, NT, OE, SA and DDC, contributed to the conception and design of the study and revision of the manuscript. All authors read and approved the final manuscript.
Author information
This work was presented in part at the annual congress of the Société de Réanimation de Langue Française (SRLF) held in January 2010, Paris, France.
This study was funded by an academic unrestricted grant (Appel d'Offre Interne) from the Caen Côte de Nacre University hospital.